ClassificationData ClassificationData::getBootstrappedDataset(UINT numSamples,bool balanceDataset) const{
Random rand;
ClassificationData newDataset;
newDataset.setNumDimensions( getNumDimensions() );
newDataset.setAllowNullGestureClass( allowNullGestureClass );
newDataset.setExternalRanges( externalRanges, useExternalRanges );
if( numSamples == 0 ) numSamples = totalNumSamples;
newDataset.reserve( numSamples );
const UINT K = getNumClasses();
//Add all the class labels to the new dataset to ensure the dataset has a list of all the labels
for(UINT k=0; k<K; k++){
newDataset.addClass( classTracker[k].classLabel );
}
if( balanceDataset ){
//Group the class indexs
Vector< Vector< UINT > > classIndexs( K );
for(UINT i=0; i<totalNumSamples; i++){
classIndexs[ getClassLabelIndexValue( data[i].getClassLabel() ) ].push_back( i );
}
//Get the class with the minimum number of examples
UINT numSamplesPerClass = (UINT)floor( numSamples / Float(K) );
//Randomly select the training samples from each class
UINT classIndex = 0;
UINT classCounter = 0;
UINT randomIndex = 0;
for(UINT i=0; i<numSamples; i++){
randomIndex = rand.getRandomNumberInt(0, (UINT)classIndexs[ classIndex ].size() );
randomIndex = classIndexs[ classIndex ][ randomIndex ];
newDataset.addSample(data[ randomIndex ].getClassLabel(), data[ randomIndex ].getSample());
if( classCounter++ >= numSamplesPerClass && classIndex+1 < K ){
classCounter = 0;
classIndex++;
}
}
}else{
//Randomly select the training samples to add to the new data set
UINT randomIndex;
for(UINT i=0; i<numSamples; i++){
randomIndex = rand.getRandomNumberInt(0, totalNumSamples);
newDataset.addSample( data[randomIndex].getClassLabel(), data[randomIndex].getSample() );
}
}
//Sort the class labels so they are in order
newDataset.sortClassLabels();
return newDataset;
}
MatrixFloat ClassificationData::getClassMean() const{
MatrixFloat mean(getNumClasses(),numDimensions);
VectorFloat counter(getNumClasses(),0);
mean.setAllValues( 0 );
for(UINT i=0; i<totalNumSamples; i++){
UINT classIndex = getClassLabelIndexValue( data[i].getClassLabel() );
for(UINT j=0; j<numDimensions; j++){
mean[classIndex][j] += data[i][j];
}
counter[ classIndex ]++;
}
for(UINT k=0; k<getNumClasses(); k++){
for(UINT j=0; j<numDimensions; j++){
mean[k][j] = counter[k] > 0 ? mean[k][j]/counter[k] : 0;
}
}
return mean;
}
MatrixFloat ClassificationData::getClassStdDev() const{
MatrixFloat mean = getClassMean();
MatrixFloat stdDev(getNumClasses(),numDimensions);
VectorFloat counter(getNumClasses(),0);
stdDev.setAllValues( 0 );
for(UINT i=0; i<totalNumSamples; i++){
UINT classIndex = getClassLabelIndexValue( data[i].getClassLabel() );
for(UINT j=0; j<numDimensions; j++){
stdDev[classIndex][j] += SQR(data[i][j]-mean[classIndex][j]);
}
counter[ classIndex ]++;
}
for(UINT k=0; k<getNumClasses(); k++){
for(UINT j=0; j<numDimensions; j++){
stdDev[k][j] = sqrt( stdDev[k][j] / Float(counter[k]-1) );
}
}
return stdDev;
}
bool ParticleClassifier::predict_( VectorDouble &inputVector ){
if( !trained ){
errorLog << "predict_(VectorDouble &inputVector) - The model has not been trained!" << endl;
return false;
}
if( numInputDimensions != inputVector.size() ){
errorLog << "predict_(VectorDouble &inputVector) - The number of features in the model " << numInputDimensions << " does not match that of the input vector " << inputVector.size() << endl;
return false;
}
//Scale the input data if needed
if( useScaling ){
for(unsigned int j=0; j<numInputDimensions; j++){
inputVector[j] = scale(inputVector[j],ranges[j].minValue,ranges[j].maxValue,0,1);
}
}
predictedClassLabel = 0;
maxLikelihood = 0;
std::fill(classLikelihoods.begin(),classLikelihoods.end(),0);
std::fill(classDistances.begin(),classDistances.end(),0);
//Update the particle filter
particleFilter.filter( inputVector );
//Count the number of particles per class
unsigned int gestureTemplate = 0;
unsigned int gestureLabel = 0;
unsigned int gestureIndex = 0;
for(unsigned int i=0; i<numParticles; i++){
gestureTemplate = (unsigned int)particleFilter[i].x[0]; //The first element in the state vector is the gesture template index
gestureLabel = particleFilter.gestureTemplates[ gestureTemplate ].classLabel;
gestureIndex = getClassLabelIndexValue( gestureLabel );
classDistances[ gestureIndex ] += particleFilter[i].w;
}
bool rejectPrediction = false;
if( useNullRejection ){
if( particleFilter.getWeightSum() < 1.0e-5 ){
rejectPrediction = true;
}
}
//Compute the class likelihoods
for(unsigned int i=0; i<numClasses; i++){
classLikelihoods[ i ] = rejectPrediction ? 0 : classDistances[i];
if( classLikelihoods[i] > maxLikelihood ){
predictedClassLabel = classLabels[i];
maxLikelihood = classLikelihoods[i];
}
}
//Estimate the phase
phase = particleFilter.getStateEstimation()[1]; //The 2nd element in the state vector is the estimatied phase
return true;
}
bool DecisionTreeClusterNode::computeError( const ClassificationData &trainingData, MatrixFloat &data, const Vector< UINT > &classLabels, Vector< MinMax > ranges, Vector< UINT > groupIndex, const UINT featureIndex, Float &threshold, Float &error ){
error = 0;
threshold = 0;
const UINT M = trainingData.getNumSamples();
const UINT K = (UINT)classLabels.size();
Float giniIndexL = 0;
Float giniIndexR = 0;
Float weightL = 0;
Float weightR = 0;
VectorFloat groupCounter(2,0);
MatrixFloat classProbabilities(K,2);
//Use this data to train a KMeans cluster with 2 clusters
KMeans kmeans;
kmeans.setNumClusters( 2 );
kmeans.setComputeTheta( true );
kmeans.setMinChange( 1.0e-5 );
kmeans.setMinNumEpochs( 1 );
kmeans.setMaxNumEpochs( 100 );
//Disable the logging to clean things up
kmeans.setTrainingLoggingEnabled( false );
if( !kmeans.train_( data ) ){
errorLog << __GRT_LOG__ << " Failed to train KMeans model for feature: " << featureIndex << std::endl;
return false;
}
//Set the split threshold as the mid point between the two clusters
const MatrixFloat &clusters = kmeans.getClusters();
threshold = 0;
for(UINT i=0; i<clusters.getNumRows(); i++){
threshold += clusters[i][0];
}
threshold /= clusters.getNumRows();
//Iterate over each sample and work out if it should be in the lhs (0) or rhs (1) group based on the current threshold
groupCounter[0] = groupCounter[1] = 0;
classProbabilities.setAllValues(0);
for(UINT i=0; i<M; i++){
groupIndex[i] = trainingData[ i ][ featureIndex ] >= threshold ? 1 : 0;
groupCounter[ groupIndex[i] ]++;
classProbabilities[ getClassLabelIndexValue(trainingData[i].getClassLabel(),classLabels) ][ groupIndex[i] ]++;
}
//Compute the class probabilities for the lhs group and rhs group
for(UINT k=0; k<K; k++){
classProbabilities[k][0] = groupCounter[0]>0 ? classProbabilities[k][0]/groupCounter[0] : 0;
classProbabilities[k][1] = groupCounter[1]>0 ? classProbabilities[k][1]/groupCounter[1] : 0;
}
//Compute the Gini index for the lhs and rhs groups
giniIndexL = giniIndexR = 0;
for(UINT k=0; k<K; k++){
giniIndexL += classProbabilities[k][0] * (1.0-classProbabilities[k][0]);
giniIndexR += classProbabilities[k][1] * (1.0-classProbabilities[k][1]);
}
weightL = groupCounter[0]/M;
weightR = groupCounter[1]/M;
error = (giniIndexL*weightL) + (giniIndexR*weightR);
return true;
}
bool TimeSeriesClassificationData::spiltDataIntoKFolds(const UINT K,const bool useStratifiedSampling){
crossValidationSetup = false;
crossValidationIndexs.clear();
//K can not be zero
if( K == 0 ){
errorLog << "spiltDataIntoKFolds(UINT K) - K can not be zero!" << std::endl;
return false;
}
//K can not be larger than the number of examples
if( K > totalNumSamples ){
errorLog << "spiltDataIntoKFolds(UINT K,bool useStratifiedSampling) - K can not be larger than the total number of samples in the dataset!" << std::endl;
return false;
}
//K can not be larger than the number of examples in a specific class if the stratified sampling option is true
if( useStratifiedSampling ){
for(UINT c=0; c<classTracker.size(); c++){
if( K > classTracker[c].counter ){
errorLog << "spiltDataIntoKFolds(UINT K,bool useStratifiedSampling) - K can not be larger than the number of samples in any given class!" << std::endl;
return false;
}
}
}
//Setup the dataset for k-fold cross validation
kFoldValue = K;
Vector< UINT > indexs( totalNumSamples );
//Work out how many samples are in each fold, the last fold might have more samples than the others
UINT numSamplesPerFold = (UINT) floor( totalNumSamples/Float(K) );
//Resize the cross validation indexs buffer
crossValidationIndexs.resize( K );
//Create the random partion indexs
Random random;
UINT randomIndex = 0;
if( useStratifiedSampling ){
//Break the data into seperate classes
Vector< Vector< UINT > > classData( getNumClasses() );
//Add the indexs to their respective classes
for(UINT i=0; i<totalNumSamples; i++){
classData[ getClassLabelIndexValue( data[i].getClassLabel() ) ].push_back( i );
}
//Randomize the order of the indexs in each of the class index buffers
for(UINT c=0; c<getNumClasses(); c++){
UINT numSamples = (UINT)classData[c].size();
for(UINT x=0; x<numSamples; x++){
//Pick a random index
randomIndex = random.getRandomNumberInt(0,numSamples);
//Swap the indexs
SWAP( classData[c][ x ] , classData[c][ randomIndex ] );
}
}
//Loop over each of the classes and add the data equally to each of the k folds until there is no data left
Vector< UINT >::iterator iter;
for(UINT c=0; c<getNumClasses(); c++){
iter = classData[ c ].begin();
UINT k = 0;
while( iter != classData[c].end() ){
crossValidationIndexs[ k ].push_back( *iter );
iter++;
k++;
k = k % K;
}
}
}else{
//Randomize the order of the data
for(UINT i=0; i<totalNumSamples; i++) indexs[i] = i;
for(UINT x=0; x<totalNumSamples; x++){
//Pick a random index
randomIndex = random.getRandomNumberInt(0,totalNumSamples);
//Swap the indexs
SWAP( indexs[ x ] , indexs[ randomIndex ] );
}
UINT counter = 0;
UINT foldIndex = 0;
for(UINT i=0; i<totalNumSamples; i++){
//Add the index to the current fold
crossValidationIndexs[ foldIndex ].push_back( indexs[i] );
//Move to the next fold if ready
if( ++counter == numSamplesPerFold && foldIndex < K-1 ){
foldIndex++;
counter = 0;
}
}
}
crossValidationSetup = true;
return true;
}
TimeSeriesClassificationData TimeSeriesClassificationData::split(const UINT trainingSizePercentage,const bool useStratifiedSampling){
//Partitions the dataset into a training dataset (which is kept by this instance of the TimeSeriesClassificationData) and
//a testing/validation dataset (which is return as a new instance of the TimeSeriesClassificationData). The trainingSizePercentage
//therefore sets the size of the data which remains in this instance and the remaining percentage of data is then added to
//the testing/validation dataset
//The dataset has changed so flag that any previous cross validation setup will now not work
crossValidationSetup = false;
crossValidationIndexs.clear();
TimeSeriesClassificationData trainingSet(numDimensions);
TimeSeriesClassificationData testSet(numDimensions);
trainingSet.setAllowNullGestureClass(allowNullGestureClass);
testSet.setAllowNullGestureClass(allowNullGestureClass);
Vector< UINT > indexs( totalNumSamples );
//Create the random partion indexs
Random random;
UINT randomIndex = 0;
if( useStratifiedSampling ){
//Break the data into seperate classes
Vector< Vector< UINT > > classData( getNumClasses() );
//Add the indexs to their respective classes
for(UINT i=0; i<totalNumSamples; i++){
classData[ getClassLabelIndexValue( data[i].getClassLabel() ) ].push_back( i );
}
//Randomize the order of the indexs in each of the class index buffers
for(UINT k=0; k<getNumClasses(); k++){
UINT numSamples = (UINT)classData[k].size();
for(UINT x=0; x<numSamples; x++){
//Pick a random index
randomIndex = random.getRandomNumberInt(0,numSamples);
//Swap the indexs
SWAP( classData[k][ x ] ,classData[k][ randomIndex ] );
}
}
//Loop over each class and add the data to the trainingSet and testSet
for(UINT k=0; k<getNumClasses(); k++){
UINT numTrainingExamples = (UINT) floor( Float(classData[k].size()) / 100.0 * Float(trainingSizePercentage) );
//Add the data to the training and test sets
for(UINT i=0; i<numTrainingExamples; i++){
trainingSet.addSample( data[ classData[k][i] ].getClassLabel(), data[ classData[k][i] ].getData() );
}
for(UINT i=numTrainingExamples; i<classData[k].size(); i++){
testSet.addSample( data[ classData[k][i] ].getClassLabel(), data[ classData[k][i] ].getData() );
}
}
//Overwrite the training data in this instance with the training data of the trainingSet
data = trainingSet.getClassificationData();
totalNumSamples = trainingSet.getNumSamples();
}else{
const UINT numTrainingExamples = (UINT) floor( Float(totalNumSamples) / 100.0 * Float(trainingSizePercentage) );
//Create the random partion indexs
Random random;
for(UINT i=0; i<totalNumSamples; i++) indexs[i] = i;
for(UINT x=0; x<totalNumSamples; x++){
//Pick a random index
randomIndex = random.getRandomNumberInt(0,totalNumSamples);
//Swap the indexs
SWAP( indexs[ x ] , indexs[ randomIndex ] );
}
//Add the data to the training and test sets
for(UINT i=0; i<numTrainingExamples; i++){
trainingSet.addSample( data[ indexs[i] ].getClassLabel(), data[ indexs[i] ].getData() );
}
for(UINT i=numTrainingExamples; i<totalNumSamples; i++){
testSet.addSample( data[ indexs[i] ].getClassLabel(), data[ indexs[i] ].getData() );
}
//Overwrite the training data in this instance with the training data of the trainingSet
data = trainingSet.getClassificationData();
totalNumSamples = trainingSet.getNumSamples();
}
return testSet;
}
bool BAG::predict_(VectorDouble &inputVector){
if( !trained ){
errorLog << "predict_(VectorDouble &inputVector) - Model Not Trained!" << endl;
return false;
}
predictedClassLabel = 0;
maxLikelihood = -10000;
if( !trained ) return false;
if( inputVector.size() != numInputDimensions ){
errorLog << "predict_(VectorDouble &inputVector) - The size of the input vector (" << inputVector.size() << ") does not match the num features in the model (" << numInputDimensions << endl;
return false;
}
if( useScaling ){
for(UINT n=0; n<numInputDimensions; n++){
inputVector[n] = scale(inputVector[n], ranges[n].minValue, ranges[n].maxValue, 0, 1);
}
}
if( classLikelihoods.size() != numClasses ) classLikelihoods.resize(numClasses);
if( classDistances.size() != numClasses ) classDistances.resize(numClasses);
//Reset the likelihoods and distances
for(UINT k=0; k<numClasses; k++){
classLikelihoods[k] = 0;
classDistances[k] = 0;
}
//Run the prediction for each classifier
double sum = 0;
UINT ensembleSize = (UINT)ensemble.size();
for(UINT i=0; i<ensembleSize; i++){
if( !ensemble[i]->predict(inputVector) ){
errorLog << "predict_(VectorDouble &inputVector) - The " << i << " classifier in the ensemble failed prediction!" << endl;
return false;
}
classLikelihoods[ getClassLabelIndexValue( ensemble[i]->getPredictedClassLabel() ) ] += weights[i];
classDistances[ getClassLabelIndexValue( ensemble[i]->getPredictedClassLabel() ) ] += ensemble[i]->getMaximumLikelihood() * weights[i];
sum += weights[i];
}
//Set the predicted class label as the most common class
double maxCount = 0;
UINT maxIndex = 0;
for(UINT i=0; i<numClasses; i++){
if( classLikelihoods[i] > maxCount ){
maxIndex = i;
maxCount = classLikelihoods[i];
}
classLikelihoods[i] /= sum;
classDistances[i] /= double(ensembleSize);
}
predictedClassLabel = classLabels[ maxIndex ];
maxLikelihood = classLikelihoods[ maxIndex ];
return true;
}
ClassificationData ClassificationData::split(const UINT trainingSizePercentage,const bool useStratifiedSampling){
//Partitions the dataset into a training dataset (which is kept by this instance of the ClassificationData) and
//a testing/validation dataset (which is return as a new instance of the ClassificationData). The trainingSizePercentage
//therefore sets the size of the data which remains in this instance and the remaining percentage of data is then added to
//the testing/validation dataset
//The dataset has changed so flag that any previous cross validation setup will now not work
crossValidationSetup = false;
crossValidationIndexs.clear();
ClassificationData trainingSet(numDimensions);
ClassificationData testSet(numDimensions);
trainingSet.setAllowNullGestureClass( allowNullGestureClass );
testSet.setAllowNullGestureClass( allowNullGestureClass );
//Create the random partion indexs
Random random;
UINT randomIndex = 0;
UINT K = getNumClasses();
if( useStratifiedSampling ){
//Break the data into seperate classes
Vector< Vector< UINT > > classData( K );
//Add the indexs to their respective classes
for(UINT i=0; i<totalNumSamples; i++){
classData[ getClassLabelIndexValue( data[i].getClassLabel() ) ].push_back( i );
}
//Randomize the order of the indexs in each of the class index buffers
for(UINT k=0; k<K; k++){
std::random_shuffle(classData[k].begin(), classData[k].end());
}
//Reserve the memory
UINT numTrainingSamples = 0;
UINT numTestSamples = 0;
for(UINT k=0; k<K; k++){
UINT numTrainingExamples = (UINT) floor( Float(classData[k].size()) / 100.0 * Float(trainingSizePercentage) );
UINT numTestExamples = ((UINT)classData[k].size())-numTrainingExamples;
numTrainingSamples += numTrainingExamples;
numTestSamples += numTestExamples;
}
trainingSet.reserve( numTrainingSamples );
testSet.reserve( numTestSamples );
//Loop over each class and add the data to the trainingSet and testSet
for(UINT k=0; k<K; k++){
UINT numTrainingExamples = (UINT) floor( Float(classData[k].getSize()) / 100.0 * Float(trainingSizePercentage) );
//Add the data to the training and test sets
for(UINT i=0; i<numTrainingExamples; i++){
trainingSet.addSample( data[ classData[k][i] ].getClassLabel(), data[ classData[k][i] ].getSample() );
}
for(UINT i=numTrainingExamples; i<classData[k].getSize(); i++){
testSet.addSample( data[ classData[k][i] ].getClassLabel(), data[ classData[k][i] ].getSample() );
}
}
}else{
const UINT numTrainingExamples = (UINT) floor( Float(totalNumSamples) / 100.0 * Float(trainingSizePercentage) );
//Create the random partion indexs
UINT randomIndex = 0;
Vector< UINT > indexs( totalNumSamples );
for(UINT i=0; i<totalNumSamples; i++) indexs[i] = i;
std::random_shuffle(indexs.begin(), indexs.end());
//Reserve the memory
trainingSet.reserve( numTrainingExamples );
testSet.reserve( totalNumSamples-numTrainingExamples );
//Add the data to the training and test sets
for(UINT i=0; i<numTrainingExamples; i++){
trainingSet.addSample( data[ indexs[i] ].getClassLabel(), data[ indexs[i] ].getSample() );
}
for(UINT i=numTrainingExamples; i<totalNumSamples; i++){
testSet.addSample( data[ indexs[i] ].getClassLabel(), data[ indexs[i] ].getSample() );
}
}
//Overwrite the training data in this instance with the training data of the trainingSet
*this = trainingSet;
//Sort the class labels in this dataset
sortClassLabels();
//Sort the class labels of the test dataset
testSet.sortClassLabels();
return testSet;
}
bool DecisionTreeThresholdNode::computeBestSplitBestRandomSplit( const UINT &numSplittingSteps, const ClassificationData &trainingData, const Vector< UINT > &features, const Vector< UINT > &classLabels, UINT &featureIndex, Float &minError ){
const UINT M = trainingData.getNumSamples();
const UINT N = (UINT)features.size();
const UINT K = (UINT)classLabels.size();
if( N == 0 ) return false;
minError = grt_numeric_limits< Float >::max();
UINT bestFeatureIndex = 0;
Float bestThreshold = 0;
Float error = 0;
Float giniIndexL = 0;
Float giniIndexR = 0;
Float weightL = 0;
Float weightR = 0;
Random random;
Vector< UINT > groupIndex(M);
VectorFloat groupCounter(2,0);
MatrixFloat classProbabilities(K,2);
//Loop over each feature and try and find the best split point
UINT m,n;
const UINT numFeatures = features.getSize();
for(m=0; m<numSplittingSteps; m++){
//Chose a random feature
n = random.getRandomNumberInt(0,numFeatures);
featureIndex = features[n];
//Randomly choose the threshold, the threshold is based on a randomly selected sample with some random scaling
threshold = trainingData[ random.getRandomNumberInt(0,M) ][ featureIndex ] * random.getRandomNumberUniform(0.8,1.2);
//Iterate over each sample and work out if it should be in the lhs (0) or rhs (1) group
groupCounter[0] = groupCounter[1] = 0;
classProbabilities.setAllValues(0);
for(UINT i=0; i<M; i++){
groupIndex[i] = trainingData[ i ][ featureIndex ] >= threshold ? 1 : 0;
groupCounter[ groupIndex[i] ]++;
classProbabilities[ getClassLabelIndexValue(trainingData[i].getClassLabel(),classLabels) ][ groupIndex[i] ]++;
}
//Compute the class probabilities for the lhs group and rhs group
for(UINT k=0; k<K; k++){
classProbabilities[k][0] = groupCounter[0]>0 ? classProbabilities[k][0]/groupCounter[0] : 0;
classProbabilities[k][1] = groupCounter[1]>0 ? classProbabilities[k][1]/groupCounter[1] : 0;
}
//Compute the Gini index for the lhs and rhs groups
giniIndexL = giniIndexR = 0;
for(UINT k=0; k<K; k++){
giniIndexL += classProbabilities[k][0] * (1.0-classProbabilities[k][0]);
giniIndexR += classProbabilities[k][1] * (1.0-classProbabilities[k][1]);
}
weightL = groupCounter[0]/M;
weightR = groupCounter[1]/M;
error = (giniIndexL*weightL) + (giniIndexR*weightR);
//Store the best threshold and feature index
if( error < minError ){
minError = error;
bestThreshold = threshold;
bestFeatureIndex = featureIndex;
}
}
//Set the best feature index that will be returned to the DecisionTree that called this function
featureIndex = bestFeatureIndex;
//Store the node size, feature index, best threshold and class probabilities for this node
set(M,featureIndex,bestThreshold,trainingData.getClassProbabilities(classLabels));
return true;
}
bool DecisionTreeThresholdNode::computeBestSplitBestIterativeSplit( const UINT &numSplittingSteps, const ClassificationData &trainingData, const Vector< UINT > &features, const Vector< UINT > &classLabels, UINT &featureIndex, Float &minError ){
const UINT M = trainingData.getNumSamples();
const UINT N = features.getSize();
const UINT K = classLabels.getSize();
if( N == 0 ) return false;
minError = grt_numeric_limits< Float >::max();
UINT bestFeatureIndex = 0;
Float bestThreshold = 0;
Float error = 0;
Float minRange = 0;
Float maxRange = 0;
Float step = 0;
Float giniIndexL = 0;
Float giniIndexR = 0;
Float weightL = 0;
Float weightR = 0;
Vector< UINT > groupIndex(M);
VectorFloat groupCounter(2,0);
Vector< MinMax > ranges = trainingData.getRanges();
MatrixFloat classProbabilities(K,2);
//Loop over each feature and try and find the best split point
for(UINT n=0; n<N; n++){
minRange = ranges[n].minValue;
maxRange = ranges[n].maxValue;
step = (maxRange-minRange)/Float(numSplittingSteps);
threshold = minRange;
featureIndex = features[n];
while( threshold <= maxRange ){
//Iterate over each sample and work out if it should be in the lhs (0) or rhs (1) group
groupCounter[0] = groupCounter[1] = 0;
classProbabilities.setAllValues(0);
for(UINT i=0; i<M; i++){
groupIndex[i] = trainingData[ i ][ featureIndex ] >= threshold ? 1 : 0;
groupCounter[ groupIndex[i] ]++;
classProbabilities[ getClassLabelIndexValue(trainingData[i].getClassLabel(),classLabels) ][ groupIndex[i] ]++;
}
//Compute the class probabilities for the lhs group and rhs group
for(UINT k=0; k<K; k++){
classProbabilities[k][0] = groupCounter[0]>0 ? classProbabilities[k][0]/groupCounter[0] : 0;
classProbabilities[k][1] = groupCounter[1]>0 ? classProbabilities[k][1]/groupCounter[1] : 0;
}
//Compute the Gini index for the lhs and rhs groups
giniIndexL = giniIndexR = 0;
for(UINT k=0; k<K; k++){
giniIndexL += classProbabilities[k][0] * (1.0-classProbabilities[k][0]);
giniIndexR += classProbabilities[k][1] * (1.0-classProbabilities[k][1]);
}
weightL = groupCounter[0]/M;
weightR = groupCounter[1]/M;
error = (giniIndexL*weightL) + (giniIndexR*weightR);
//Store the best threshold and feature index
if( error < minError ){
minError = error;
bestThreshold = threshold;
bestFeatureIndex = featureIndex;
}
//Update the threshold
threshold += step;
}
}
//Set the best feature index that will be returned to the DecisionTree that called this function
featureIndex = bestFeatureIndex;
//Store the node size, feature index, best threshold and class probabilities for this node
set(M,featureIndex,bestThreshold,trainingData.getClassProbabilities(classLabels));
return true;
}
